When Platform Conversions, GA4, and CRM Disagree
Northwind Analytics is a B2B SaaS company with a 21-day sales cycle. Last quarter its CRM recorded 420 marketing-qualified leads, while the primary ad platform reported 610 conversions and GA4 reported 300 key events. The team runs search, LinkedIn, and retargeting campaigns across five regions. Consent Mode v2 was enabled eight weeks ago; the consent rate is 62%, and EU traffic is 18% of sessions. The ad platform uses a 30-day click, 1-day view window and counts form starts as conversions, while GA4 counts only form submissions. The CRM is the system of record for sales-accepted leads. A budget review is scheduled for Monday, and the CMO wants to shift 25% of spend to the best-performing channel. The analytics lead has two days and no engineering support. What should the team do?
1. Explain the likely sources of the 610 vs 420 vs 300 gap and rank which number is most defensible for budget decisions.
Read the worked answer
The three figures measure different things, so none is automatically wrong.
- Conversion definition: the ad platform counts form starts; GA4 counts form submissions; the CRM counts sales-accepted MQLs. A form start is a weaker event, so 610 should be highest. A submission can still fail validation or be a duplicate, so GA4's 300 may be lower than 420.
- Attribution window and modeling: 30-day click plus 1-day view gives the platform credit for conversions GA4 may assign to direct or organic, especially with a 21-day sales cycle. The platform also models conversions when consent or identifiers are missing, inflating reported totals.
- Consent and identity: Consent Mode v2 at 62% consent means 38% of traffic is modeled or cookieless. GA4 will often undercount relative to the CRM when a lead later identifies through a sales call or form.
- CRM timing: a 21-day cycle means last quarter's spend creates leads that may be accepted this quarter. Pulling CRM by created date versus accepted date changes 420.
Use CRM MQL count as the north star for spend allocation, but only after matching each lead to a source and campaign and normalizing by spend. Platform conversions are useful for in-platform optimization, not for cross-channel budget shifts. GA4 is useful for on-site behavior and funnel diagnostics, not as the revenue denominator.
Decision rule: compare cost per CRM-accepted MQL and MQL-to-opportunity rate by channel over matched cohorts. If the platform-to-CRM gap is stable by channel, rank on platform numbers with a correction factor; if the gap varies by channel, use CRM. Do not shift 25% on raw platform conversions.
2. Design a two-week measurement triage plan with no engineering support.
Read the worked answer
Days 1-2 - define and reconcile. Write a one-page metric dictionary: form start, form submit, MQL, accepted lead, opportunity. Export CRM leads by created date, source, campaign, and region for the last two quarters. Check UTM coverage: what share of CRM leads have a campaign value? If coverage is below 80%, channel comparison is unreliable.
Days 3-4 - quantify the gap. Build a pivot of platform conversions, GA4 key events, and CRM MQLs by channel. Calculate a platform-to-CRM ratio per channel and its variance. Check consent impact by region by comparing EU and non-EU conversion rates. Look for duplicates and test submissions.
Days 5-6 - inspect the funnel. For one high-spend channel, trace 20-30 leads from click ID or UTM to CRM. Note where they drop: form start without submit, submit without CRM record, CRM record without accepted status.
Days 7-8 - set a provisional correction. Apply the observed channel-level ratio to platform numbers as a temporary estimate, with a range. Flag channels where the ratio is unstable.
Days 9-10 - recommend a decision rule. Use cost per CRM MQL and MQL-to-opportunity, not raw conversions. Propose a 10% budget shift into a geo or channel holdout for four weeks rather than a 25% reallocation.
Week 2 - validate and document. Run a small holdout if possible, document assumptions, and list engineering fixes needed: consent-safe conversion tracking, CRM-campaign mapping, and deduplication. Deliver a dashboard with confidence caveats.
3. What recommendation should the analyst give the CMO on Monday?
Read the worked answer
Do not shift 25% of spend based on the platform's 610 conversions. That number is inflated by form-start definitions, view-through attribution, and modeling. The 420 CRM MQLs are the most defensible denominator, but they are not yet matched cleanly to campaigns.
Give the CMO a conditional recommendation:
- Freeze 25% reallocation for two weeks. Apply a provisional channel-level correction to platform conversions using the observed platform-to-CRM ratio. Rank channels by cost per CRM MQL, not cost per platform conversion.
- Shift 10% now only if the corrected data is consistent. For example, if search cost per MQL is $180 and LinkedIn is $310, and MQL-to-opportunity rates are similar, a 10% shift toward search is defensible. Keep 15% in reserve.
- Run a four-week holdout or geo test on the proposed winner before moving the remaining 15%. Define success as lower cost per accepted MQL without a drop in MQL-to-opportunity rate.
- Fix the measurement debt: require CRM campaign mapping, deduplication, and a consent-safe conversion definition before next quarter's budget cycle.
The message should be: the direction may be right, but the evidence is not strong enough for a 25% swing. A smaller shift plus a test protects the budget review and produces a defensible answer.